A Multidimensional Model for Correct Aggregation of Geographic Measures - Evolving Application Domains of Data Warehousing and Mining - page 180

Mode. To complete our case study, we have to
recall that no aggregation function is provided to
get the value of the attribute representing the area
of the damaged zone. Indeed, the area is a derived
measure calculated using the geometry resulting
from the spatial aggregation of the component
geographic objects.

alphanumeric data. The model supports overlap-
ping geometries and aggregation constraints as
well as our 3-step framework. However, it sup-
ports only partially the dependency of spatial
and alphanumeric functions as it does not allow
disaggregation functions. We have addressed this
issue in this paper
Currently, we are working on the implemen-
tation of our framework in the SOLAP system
GeWOlap (Bimonte et al., 2006b). We are imple-
menting the three steps of our framework using
PL/SQL functions and user-defined aggregation
functions provided by Oracle. Semantics of mea-
sures (valid aggregation functions) are defined
in XML files representing the multidimensional
application in the ROLAP Server Mondrian.
Semantics of measures have to be parsed before
any OLAP query computation in order to verify
dependencies between spatial and alphanumeric
functions. The main issues of this approach deal
with performances in large spatial data ware-
houses. Developing ad-hoc pre-aggregation and
indexing techniques for geographic measures are
our future work.

concluSIon And dIScuSSIon

Spatial OLAP refers to the integration of spatial
data in OLAP. Correct aggregation is crucial
in multidimensional analysis. In this paper, we
provide an overview of solutions for aggrega-
tion of geographic objects in multidimensional,
geostatistic, GIS and SOLAP models. By intro-
ducing a case study concerning the monitoring of
natural phenomena, we define the aggregation of
geographic measures as a -three-step process in
which two constraints on aggregation functions are
defined. They extend classical OLAP aggregation
constraints with dependence between spatial and
alphanumeric aggregation functions applied to the
attributes of geographic measures. We present an
extension of the logical multidimensional model
GeoCube (Bimonte et al., 2006), which formal-
izes our approach. Alphanumeric aggregation
functions are decomposed into two aggregation
functions and two constraints are applied to them to
take into account the additivity of the measures and
the dependency of spatial and alphanumeric ag-
gregation functions. By this way, GeoCube ensures
the correct aggregation of geographic measures.
This requirement is not supported by existing
spatio-multidimensional model. Indeed, most of
SOLAP models do not provide any aggregation
constraint for aggregation functions. Jensen et al.
(2004) define aggregation constraints, but they
reduce spatial measures to classical data without
taking into account the spatial component of the
geographic information. Finally, only Pedersen
& Tryfona (2001) provide a model that explicitly
supports aggregation constraints for spatial and

AcknoWledgMent

Authors wish to thank Pr. Pierre Dumolard for
precious discussions and material on GIS and
geostatistics.

referenceS

Abelló, A., Samos, J., & Saltor, F. (2006). YAM2:
a multidimensional conceptual model extending
UML. Information Systems , 31 (6), 541-567.
doi:10.1016/j.is.2004.12.002

Ahmed, T., & Miquel, M. (2005). Multidimen-
sional Structures Dedicated to Continuous Spa-
tiotemporal Phenomena. In Proceedings of the
22th BNCOD (pp. 29-40). Berlin: Springer.